基于ConvLSTM和PredRNN的大气能见度预报方法

大气与环境光学学报 ›› 2023, Vol. 18 ›› Issue (5): 434-444.

基于ConvLSTM和PredRNN的大气能见度预报方法

包旭伦 1, 李毅 2*, 胡译文 2,3, 王阳 4, 牛丹 5, 臧增亮 2, 陈夕松 5

1 东南大学软件学院, 江苏苏州 215123; 2 国防科技大学气象海洋学院, 湖南长沙 410000; 3 南京信息工程大学大气科学学院, 江苏南京 210044; 4 北京弘象科技有限公司, 北京 100089; 5 东南大学自动化学院, 江苏南京 211189

收稿日期:2022-02-11 修回日期:2022-04-26 出版日期:2023-09-28 发布日期:2023-10-11
通讯作者: E-mail: liyiqxxy@163.com E-mail:liyiqxxy@163.com
作者简介:包旭伦(1998- ), 内蒙古赤峰人, 硕士研究生, 主要从事深度学习应用气象预测方面的研究。E-mail: 220205610@seu.edu.cn
基金资助:
国家自然科学基金 (41975167, 41775123)

Atmospheric visibility prediction method based on ConvLSTM and PredRNN

BAO Xulun 1, LI Yi 2*, Hu Yiwen 2,3, WANG Yang 4, NIU Dan 5, ZANG Zengliang 2, CHEN Xisong 5

1 School of Software, Southeast University, Suzhou 215123, China; 2 College of Meteorology and Oceanography, National University of Defense Technology, Changsha 410000, China; 3 College of Atmospheric Sciences, Nanjing University of Information Science & Technology, Nanjing 210044, China; 4 Beijing Hongxiang Technology Co., LTD, Beijing 100089, China; 5 School of Automation, Southeast University, Nanjing 211189, China

Received:2022-02-11 Revised:2022-04-26 Online:2023-09-28 Published:2023-10-11
Contact: Yi -LI E-mail:liyiqxxy@163.com
Supported by:
National Natural Science Foundation of China

摘要/Abstract

摘要： 精准的大气能见度预报对空气污染治理、保障公共交通安全等方面具有重要意义。基于2019 年12 月1 日至 2020 年9 月23 日国家气象信息中心观测的大气能见度站点数据，分别采用ConvLSTM模型和PredRNN模型对中国中东部地区的能见度进行12 h 预报，并对这两种模型的预报结果进行评价。试验表明， PredRNN 模型相对于经典的 ConvLSTM模型在大气能见度预报、图像质量评价指标和预报指标上都有更好的表现。此外，分析还表明，相对于 ConvLSTM模型， PredRNN模型对4000 m中等级别雾区预报效果随时间延长有明显提升。

关键词: 大气能见度预报, 预测递归神经网络, 时空预测, 提高精度

Abstract: Accurate forecast of atmospheric visibility is of great significance to air pollution control and public transportation safety. Based on the atmospheric visibility data observed by the National Meteorological Information Center from December 1, 2019 to September 23, 2020, ConvLSTM model and PredRNN model were used to forecast visibility over central and eastern China for 12 h in this work, and the forecast results of the two models were evaluated. The results show that PredRNN model performs better than the traditional ConvLSTM model in atmospheric visibility forecast, image quality evaluation index and forecast index. In addition, it is also found that compared with ConvLSTM model, PredRNN model has improved significantly in forecasting 4000 m medium-level fog area over time.

Key words: prediction of atmospheric visibility, predictive recurrent neural networks, spatiotemporal prediction, improve the accuracy

中图分类号:

P456.9

包旭伦, 李毅, 胡译文, 王阳, 牛丹, 臧增亮, 陈夕松 . 基于ConvLSTM和PredRNN的大气能见度预报方法[J]. 大气与环境光学学报, 2023, 18(5): 434-444.

BAO Xulun , LI Yi , Hu Yiwen , WANG Yang , NIU Dan , ZANG Zengliang , CHEN Xisong . Atmospheric visibility prediction method based on ConvLSTM and PredRNN[J]. Journal of Atmospheric and Environmental Optics, 2023, 18(5): 434-444.

参考文献 16

[1]	Wen J P, Lan J, Liu F. Experiment on objective forecast methods for the low visibility of Guangzhou Baiyun Airport [J].
	Advances in Meteorological Science and Technology, 2021, 11(2): 176-180.
	文俊鹏, 蓝静, 刘峰. 广州白云机场低能见度客观预报方法试验 [J]. 气象科技进展, 2021, 11(2): 176-180.
[2]	Abdellaoui I A, Mehrkanoon S. Deep multi-stations weather forecasting: Explainable recurrent convolutional neural networks
	[EB/OL]. https://arxiv.org/abs/2009.11239.
[3]	Su J X, Wang S T, Chen H M, et al. Atmospheric visibility forcast based on GA-BP neural network model [J]. Computer
	Knowledge and Technology, 2020, 16(18): 19-21.
	苏靖晰, 王圣堂, 陈红敏, 等. 基于GA-BP神经网络模型的大气能见度预测 [J]. 电脑知识与技术, 2020, 16(18): 19-21.
[4]	Shi X J, Chen Z R, Wang H, et al. Convolutional LSTM network: A machine learning approach for precipitation nowcasting
[C]	Proceedings of the 28th International Conference on Neural Information Processing Systems, December 7, 2015,
	Montreal, Canada. MIT Press, 2015: 802-810.
[5]	Guo H Y, Chen M X, Han L, et al. High resolution nowcasting experiment of severe convection based on deep learning [J].
	Acta Meteorologica Sinica, 2019, 77(4): 715-727.
	郭瀚阳, 陈明轩, 韩雷, 等. 基于深度学习的强对流高分辨率临近预报试验 [J]. 气象学报, 2019, 77(4): 715-727.
[6]	Niu D, Huang J H, Zang Z L, et al. Two-stage spatiotemporal context refinement network for precipitation nowcasting [J].
	Remote Sensing, 2021, 13(21): 4285.
[7]	Xu Z R, Wang Y B, Long M S, et al. PredCNN: Predictive learning with cascade convolutions [C]. Proceedings of the 27th
	International Joint Conference on Artificial Intelligence. July 13-19, 2018, Stockholm, Sweden. AAAI Press, 2018: 2940-2947.
[8]	Rumelhart D E, Hinton G E, Williams R J. Learning representations by back-propagating errors [J]. Nature, 1986, 323(6088):
53	3-536.
[9]	Hochreiter S, Schmidhuber J. Long short-term memory [J]. Neural Computation, 1997, 9(8): 1735-1780.
[10]	Jiang L, Wang B S, Hong L, et al. Airport visibility prediction based on CNN deep learning model [J]. China Computer &
	Communication, 2020, 32(23): 43-46.
	江亮, 王保升, 洪磊, 等. 基于CNN深度学习模型的机场能见度预测 [J]. 信息与电脑, 2020, 32(23): 43-46.
[11]	Deng T. Airport Visibility Prediction Based on LSTM Neural Network [D]. Jinan: Shandong University, 2019.
	邓拓. 基于LSTM神经网络的机场能见度预测 [D]. 济南: 山东大学, 2019.
[12]	Wang Y B, Long M S, Wang J M, et al. PredRNN: Recurrent neural networks for predictive learning using spatiotemporal
	LSTMs [C]. Proceedings of the 31st International Conference on Neural Information Processing Systems. December 4-9, 2017,
	Long Beach, California, USA. New York: ACM, 2017: 879-888.
[13]	LeCun Y, Bottou L, Bengio Y, et al. Gradient-based learning applied to document recognition [J]. Proceedings of the IEEE,
19	98, 86(11): 2278-2324.
[14]	Oh J, Guo X, Lee H, et al. Action-conditional video prediction using deep networks in Atari Games [C]. Proceedings of the
28	th International Conference on Neural Information Processing Systems. December 7, 2015, Montreal, Canada, MIT Press,
20	15: 2863-2871.
[15]	Wang Z, Bovik A C, Sheikh H R, et al. Image quality assessment: From error visibility to structural similarity [J]. IEEE
	Transactions on Image Processing: A Publication of the IEEE Signal Processing Society, 2004, 13(4): 600-612.
[16]	Zhang R, Isola P, Efros A A, et al. The unreasonable effectiveness of deep features as a perceptual metric [C]. 2018 IEEE/CVF
	Conference on Computer Vision and Pattern Recognition. June 18-23, 2018, Salt Lake City, UT, USA. IEEE, 2018: 586-595.